{ "schemaVersion": 1, "slug": "kv-separation-cleaning", "kind": "dataset", "title": "Key-value separation: cleaning-cost sensitivity", "summary": "Thirty deterministic byte-accounting cases compare inline records with separated values, including relocation-driven index writes. Invented inputs, not engine measurements.", "published": "2026-10-12", "updated": "2026-10-12", "evidenceType": "analytical-model-evaluation", "method": "For equal-sized records, K=16 key bytes, P=16 reference bytes, H=8 log-header bytes; V in {16,128,1024} bytes; C in {4,10} tree output/input-byte factors; selected-region live fraction u in {0,0.25,0.5,0.75,0.9}. Inline=C(K+V); separated=[C(K+P)+K+V+H]/(1-u). Every relocated record induces an index update charged at C. Each ratio uses K+V new logical bytes as denominator. Original steady-state accounting example, not a fit to a paper.", "sourceUrl": "https://www.usenix.org/system/files/conference/fast16/fast16-papers-lu.pdf", "humanPage": "https://hesela.com/analyses/key-value-separation-cleaning-cost/", "downloadPath": "/data/datasets/kv-separation-cleaning.csv", "codePath": "/research/kv-cleaning.mjs", "license": "CC0-1.0 for Hesela's original code and computed rows. Source papers and documentation retain their own rights.", "licenseUrl": "https://creativecommons.org/publicdomain/zero/1.0/", "columns": [ { "name": "keyBytes", "type": "number", "description": "Assumed key size, bytes." }, { "name": "valueBytes", "type": "number", "description": "Assumed value size, bytes." }, { "name": "pointerBytes", "type": "number", "description": "Assumed reference size in the separated tree, bytes." }, { "name": "logHeaderBytes", "type": "number", "description": "Assumed extra log header size, bytes." }, { "name": "treeWriteFactor", "type": "number", "description": "Tree output bytes per inserted tree-record byte, including initial table writes and rewrites; same for both layouts." }, { "name": "liveFraction", "type": "number", "description": "Live-byte fraction of the selected cleaning regions, not whole-database utilization." }, { "name": "logicalBytes", "type": "number", "description": "New logical input bytes per key-value insertion: keyBytes + valueBytes." }, { "name": "relocationFactor", "type": "number", "description": "Total append-plus-relocation multiplier: 1/(1-liveFraction)." }, { "name": "inlineBytes", "type": "number", "description": "Modeled tree bytes written per new logical insertion in the inline layout." }, { "name": "separatedTreeBytes", "type": "number", "description": "Modeled tree bytes per new insertion including pointer updates for relocated records." }, { "name": "valueLogBytes", "type": "number", "description": "Modeled log bytes per new insertion including all repeated copies and headers." }, { "name": "separatedBytes", "type": "number", "description": "Separated tree plus log bytes per new insertion; excludes WAL and lower-layer writes." }, { "name": "inlineWriteRatio", "type": "number", "description": "inlineBytes / logicalBytes, modeled byte ratio." }, { "name": "separatedWriteRatio", "type": "number", "description": "separatedBytes / logicalBytes, modeled byte ratio." }, { "name": "separatedToInline", "type": "number", "description": "separatedBytes / inlineBytes; below 1 means fewer modeled writes." }, { "name": "breakEvenLiveFraction", "type": "number|null", "description": "Selected-region live fraction at equal cost; null/empty CSV means no nonnegative crossover (separation loses even at zero live bytes)." } ], "limits": [ "All inputs are invented and fixed; no devices, traces or database engines were measured.", "Assumes steady-state cleaning with equal record sizes and a fixed live fraction of selected regions.", "Identical C for inline, separated and relocation-driven tree updates isolates layout; a real smaller index can have a different C.", "No WAL, reads, compression, caching, coalescing, metadata beyond H/P, alignment, latency or NAND writes.", "Charging an index update for every relocation does not represent compaction-integrated BlobDB reclamation.", "u=1 cannot reclaim net space and is rejected; u is not storage utilization or a fill-limit recommendation.", "No statistical confidence intervals apply to these deterministic cases; applicability remains uncertain." ], "citations": [ { "id": "wisckey", "title": "Lanyue Lu, Thanumalayan Sankaranarayana Pillai, Andrea C. Arpaci-Dusseau and Remzi H. Arpaci-Dusseau. WiscKey: Separating Keys from Values in SSD-conscious Storage. USENIX FAST 2016, pp. 133-148, sections 3.2-3.3 and 4.", "url": "https://www.usenix.org/system/files/conference/fast16/fast16-papers-lu.pdf", "accessed": "2026-10-12" }, { "id": "hashkv", "title": "Helen H. W. Chan, Yongkun Li, Patrick P. C. Lee and Yinlong Xu. HashKV: Enabling Efficient Updates in KV Storage via Hashing. USENIX ATC 2018, pp. 1007-1019, sections 3-4.", "url": "https://www.usenix.org/system/files/conference/atc18/atc18-chan.pdf", "accessed": "2026-10-12" }, { "id": "blobdb", "title": "RocksDB project. BlobDB: integrated implementation, garbage collection and configuration. Documentation reviewed 12 October 2026; page displays revision date 22 September 2022.", "url": "https://github.com/facebook/rocksdb/wiki/BlobDB", "accessed": "2026-10-12" } ], "generatedAt": "2026-10-12T03:22:51.748Z", "runtime": { "node": "v24.21.0", "dependencies": "none", "seed": "not applicable: deterministic arithmetic" }, "rows": [ { "keyBytes": 16, "valueBytes": 16, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0, "logicalBytes": 32, "relocationFactor": 1, "inlineBytes": 128, "separatedTreeBytes": 128, "valueLogBytes": 40, "separatedBytes": 168, "inlineWriteRatio": 4, "separatedWriteRatio": 5.25, "separatedToInline": 1.3125, "breakEvenLiveFraction": null }, { "keyBytes": 16, "valueBytes": 16, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.25, "logicalBytes": 32, "relocationFactor": 1.3333333333333333, "inlineBytes": 128, "separatedTreeBytes": 170.66666666666666, "valueLogBytes": 53.33333333333333, "separatedBytes": 224, "inlineWriteRatio": 4, "separatedWriteRatio": 7, "separatedToInline": 1.75, "breakEvenLiveFraction": null }, { "keyBytes": 16, "valueBytes": 16, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.5, "logicalBytes": 32, "relocationFactor": 2, "inlineBytes": 128, "separatedTreeBytes": 256, "valueLogBytes": 80, "separatedBytes": 336, "inlineWriteRatio": 4, "separatedWriteRatio": 10.5, "separatedToInline": 2.625, "breakEvenLiveFraction": null }, { "keyBytes": 16, "valueBytes": 16, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.75, "logicalBytes": 32, "relocationFactor": 4, "inlineBytes": 128, "separatedTreeBytes": 512, "valueLogBytes": 160, "separatedBytes": 672, "inlineWriteRatio": 4, "separatedWriteRatio": 21, "separatedToInline": 5.25, "breakEvenLiveFraction": null }, { "keyBytes": 16, "valueBytes": 16, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.9, "logicalBytes": 32, "relocationFactor": 10.000000000000002, "inlineBytes": 128, "separatedTreeBytes": 1280.0000000000002, "valueLogBytes": 400.00000000000006, "separatedBytes": 1680.0000000000002, "inlineWriteRatio": 4, "separatedWriteRatio": 52.50000000000001, "separatedToInline": 13.125000000000002, "breakEvenLiveFraction": null }, { "keyBytes": 16, "valueBytes": 128, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0, "logicalBytes": 144, "relocationFactor": 1, "inlineBytes": 576, "separatedTreeBytes": 128, "valueLogBytes": 152, "separatedBytes": 280, "inlineWriteRatio": 4, "separatedWriteRatio": 1.9444444444444444, "separatedToInline": 0.4861111111111111, "breakEvenLiveFraction": 0.5138888888888888 }, { "keyBytes": 16, "valueBytes": 128, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.25, "logicalBytes": 144, "relocationFactor": 1.3333333333333333, "inlineBytes": 576, "separatedTreeBytes": 170.66666666666666, "valueLogBytes": 202.66666666666666, "separatedBytes": 373.3333333333333, "inlineWriteRatio": 4, "separatedWriteRatio": 2.5925925925925926, "separatedToInline": 0.6481481481481481, "breakEvenLiveFraction": 0.5138888888888888 }, { "keyBytes": 16, "valueBytes": 128, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.5, "logicalBytes": 144, "relocationFactor": 2, "inlineBytes": 576, "separatedTreeBytes": 256, "valueLogBytes": 304, "separatedBytes": 560, "inlineWriteRatio": 4, "separatedWriteRatio": 3.888888888888889, "separatedToInline": 0.9722222222222222, "breakEvenLiveFraction": 0.5138888888888888 }, { "keyBytes": 16, "valueBytes": 128, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.75, "logicalBytes": 144, "relocationFactor": 4, "inlineBytes": 576, "separatedTreeBytes": 512, "valueLogBytes": 608, "separatedBytes": 1120, "inlineWriteRatio": 4, "separatedWriteRatio": 7.777777777777778, "separatedToInline": 1.9444444444444444, "breakEvenLiveFraction": 0.5138888888888888 }, { "keyBytes": 16, "valueBytes": 128, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.9, "logicalBytes": 144, "relocationFactor": 10.000000000000002, "inlineBytes": 576, "separatedTreeBytes": 1280.0000000000002, "valueLogBytes": 1520.0000000000002, "separatedBytes": 2800.0000000000005, "inlineWriteRatio": 4, "separatedWriteRatio": 19.444444444444446, "separatedToInline": 4.861111111111112, "breakEvenLiveFraction": 0.5138888888888888 }, { "keyBytes": 16, "valueBytes": 1024, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0, "logicalBytes": 1040, "relocationFactor": 1, "inlineBytes": 4160, "separatedTreeBytes": 128, "valueLogBytes": 1048, "separatedBytes": 1176, "inlineWriteRatio": 4, "separatedWriteRatio": 1.1307692307692307, "separatedToInline": 0.2826923076923077, "breakEvenLiveFraction": 0.7173076923076923 }, { "keyBytes": 16, "valueBytes": 1024, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.25, "logicalBytes": 1040, "relocationFactor": 1.3333333333333333, "inlineBytes": 4160, "separatedTreeBytes": 170.66666666666666, "valueLogBytes": 1397.3333333333333, "separatedBytes": 1568, "inlineWriteRatio": 4, "separatedWriteRatio": 1.5076923076923077, "separatedToInline": 0.3769230769230769, "breakEvenLiveFraction": 0.7173076923076923 }, { "keyBytes": 16, "valueBytes": 1024, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.5, "logicalBytes": 1040, "relocationFactor": 2, "inlineBytes": 4160, "separatedTreeBytes": 256, "valueLogBytes": 2096, "separatedBytes": 2352, "inlineWriteRatio": 4, "separatedWriteRatio": 2.2615384615384615, "separatedToInline": 0.5653846153846154, "breakEvenLiveFraction": 0.7173076923076923 }, { "keyBytes": 16, "valueBytes": 1024, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.75, "logicalBytes": 1040, "relocationFactor": 4, "inlineBytes": 4160, "separatedTreeBytes": 512, "valueLogBytes": 4192, "separatedBytes": 4704, "inlineWriteRatio": 4, "separatedWriteRatio": 4.523076923076923, "separatedToInline": 1.1307692307692307, "breakEvenLiveFraction": 0.7173076923076923 }, { "keyBytes": 16, "valueBytes": 1024, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 4, "liveFraction": 0.9, "logicalBytes": 1040, "relocationFactor": 10.000000000000002, "inlineBytes": 4160, "separatedTreeBytes": 1280.0000000000002, "valueLogBytes": 10480.000000000002, "separatedBytes": 11760.000000000002, "inlineWriteRatio": 4, "separatedWriteRatio": 11.30769230769231, "separatedToInline": 2.8269230769230775, "breakEvenLiveFraction": 0.7173076923076923 }, { "keyBytes": 16, "valueBytes": 16, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0, "logicalBytes": 32, "relocationFactor": 1, "inlineBytes": 320, "separatedTreeBytes": 320, "valueLogBytes": 40, "separatedBytes": 360, "inlineWriteRatio": 10, "separatedWriteRatio": 11.25, "separatedToInline": 1.125, "breakEvenLiveFraction": null }, { "keyBytes": 16, "valueBytes": 16, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.25, "logicalBytes": 32, "relocationFactor": 1.3333333333333333, "inlineBytes": 320, "separatedTreeBytes": 426.66666666666663, "valueLogBytes": 53.33333333333333, "separatedBytes": 479.99999999999994, "inlineWriteRatio": 10, "separatedWriteRatio": 14.999999999999998, "separatedToInline": 1.4999999999999998, "breakEvenLiveFraction": null }, { "keyBytes": 16, "valueBytes": 16, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.5, "logicalBytes": 32, "relocationFactor": 2, "inlineBytes": 320, "separatedTreeBytes": 640, "valueLogBytes": 80, "separatedBytes": 720, "inlineWriteRatio": 10, "separatedWriteRatio": 22.5, "separatedToInline": 2.25, "breakEvenLiveFraction": null }, { "keyBytes": 16, "valueBytes": 16, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.75, "logicalBytes": 32, "relocationFactor": 4, "inlineBytes": 320, "separatedTreeBytes": 1280, "valueLogBytes": 160, "separatedBytes": 1440, "inlineWriteRatio": 10, "separatedWriteRatio": 45, "separatedToInline": 4.5, "breakEvenLiveFraction": null }, { "keyBytes": 16, "valueBytes": 16, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.9, "logicalBytes": 32, "relocationFactor": 10.000000000000002, "inlineBytes": 320, "separatedTreeBytes": 3200.0000000000005, "valueLogBytes": 400.00000000000006, "separatedBytes": 3600.0000000000005, "inlineWriteRatio": 10, "separatedWriteRatio": 112.50000000000001, "separatedToInline": 11.250000000000002, "breakEvenLiveFraction": null }, { "keyBytes": 16, "valueBytes": 128, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0, "logicalBytes": 144, "relocationFactor": 1, "inlineBytes": 1440, "separatedTreeBytes": 320, "valueLogBytes": 152, "separatedBytes": 472, "inlineWriteRatio": 10, "separatedWriteRatio": 3.2777777777777777, "separatedToInline": 0.3277777777777778, "breakEvenLiveFraction": 0.6722222222222223 }, { "keyBytes": 16, "valueBytes": 128, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.25, "logicalBytes": 144, "relocationFactor": 1.3333333333333333, "inlineBytes": 1440, "separatedTreeBytes": 426.66666666666663, "valueLogBytes": 202.66666666666666, "separatedBytes": 629.3333333333333, "inlineWriteRatio": 10, "separatedWriteRatio": 4.37037037037037, "separatedToInline": 0.437037037037037, "breakEvenLiveFraction": 0.6722222222222223 }, { "keyBytes": 16, "valueBytes": 128, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.5, "logicalBytes": 144, "relocationFactor": 2, "inlineBytes": 1440, "separatedTreeBytes": 640, "valueLogBytes": 304, "separatedBytes": 944, "inlineWriteRatio": 10, "separatedWriteRatio": 6.555555555555555, "separatedToInline": 0.6555555555555556, "breakEvenLiveFraction": 0.6722222222222223 }, { "keyBytes": 16, "valueBytes": 128, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.75, "logicalBytes": 144, "relocationFactor": 4, "inlineBytes": 1440, "separatedTreeBytes": 1280, "valueLogBytes": 608, "separatedBytes": 1888, "inlineWriteRatio": 10, "separatedWriteRatio": 13.11111111111111, "separatedToInline": 1.3111111111111111, "breakEvenLiveFraction": 0.6722222222222223 }, { "keyBytes": 16, "valueBytes": 128, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.9, "logicalBytes": 144, "relocationFactor": 10.000000000000002, "inlineBytes": 1440, "separatedTreeBytes": 3200.0000000000005, "valueLogBytes": 1520.0000000000002, "separatedBytes": 4720.000000000001, "inlineWriteRatio": 10, "separatedWriteRatio": 32.777777777777786, "separatedToInline": 3.2777777777777786, "breakEvenLiveFraction": 0.6722222222222223 }, { "keyBytes": 16, "valueBytes": 1024, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0, "logicalBytes": 1040, "relocationFactor": 1, "inlineBytes": 10400, "separatedTreeBytes": 320, "valueLogBytes": 1048, "separatedBytes": 1368, "inlineWriteRatio": 10, "separatedWriteRatio": 1.3153846153846154, "separatedToInline": 0.13153846153846155, "breakEvenLiveFraction": 0.8684615384615384 }, { "keyBytes": 16, "valueBytes": 1024, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.25, "logicalBytes": 1040, "relocationFactor": 1.3333333333333333, "inlineBytes": 10400, "separatedTreeBytes": 426.66666666666663, "valueLogBytes": 1397.3333333333333, "separatedBytes": 1824, "inlineWriteRatio": 10, "separatedWriteRatio": 1.7538461538461538, "separatedToInline": 0.1753846153846154, "breakEvenLiveFraction": 0.8684615384615384 }, { "keyBytes": 16, "valueBytes": 1024, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.5, "logicalBytes": 1040, "relocationFactor": 2, "inlineBytes": 10400, "separatedTreeBytes": 640, "valueLogBytes": 2096, "separatedBytes": 2736, "inlineWriteRatio": 10, "separatedWriteRatio": 2.6307692307692307, "separatedToInline": 0.2630769230769231, "breakEvenLiveFraction": 0.8684615384615384 }, { "keyBytes": 16, "valueBytes": 1024, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.75, "logicalBytes": 1040, "relocationFactor": 4, "inlineBytes": 10400, "separatedTreeBytes": 1280, "valueLogBytes": 4192, "separatedBytes": 5472, "inlineWriteRatio": 10, "separatedWriteRatio": 5.2615384615384615, "separatedToInline": 0.5261538461538462, "breakEvenLiveFraction": 0.8684615384615384 }, { "keyBytes": 16, "valueBytes": 1024, "pointerBytes": 16, "logHeaderBytes": 8, "treeWriteFactor": 10, "liveFraction": 0.9, "logicalBytes": 1040, "relocationFactor": 10.000000000000002, "inlineBytes": 10400, "separatedTreeBytes": 3200.0000000000005, "valueLogBytes": 10480.000000000002, "separatedBytes": 13680.000000000002, "inlineWriteRatio": 10, "separatedWriteRatio": 13.153846153846155, "separatedToInline": 1.3153846153846156, "breakEvenLiveFraction": 0.8684615384615384 } ], "rowCount": 30, "artifacts": { "modelSha256": "422ed6005717618f92da2ad75d6190fa1de9d5fc6410eedca745383277369bef", "csvSha256": "491a64cf286280c930a5202dbb69fc6833060d29bb9de9f86319f671a309fd43" } }